Pathway to high throughput, low cost indium-free transparent electrodes † – 13899 This journal is

A roll-to-roll compatible, high throughput process is reported for the production of highly conductive, transparent planar electrode comprising an interwoven network of silver nanowires and single walled carbon nanotubes imbedded into poly(3,4-ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS). The planar electrode has a sheet resistance of between 4 and 7 U , (cid:1) 1 and a transmission of >86% between 800 and 400 nm with a ﬁ gure of merit of between 344 and 400 U (cid:1) 1 . The nanocomposite electrode is highly ﬂ exible and retains a low sheet resistance after bending at a radius of 5 mm for up to 500 times without loss. Organic photovoltaic devices containing the planar nanocomposite electrodes had e ﬃ ciencies of (cid:3) 90% of control devices that used indium tin oxide as the transparent conducting electrode.


Introduction
Roll-to-roll (R2R) manufacturing of organic photovoltaic (OPV) devices has become a key focus of research activity since it has the potential to realise the cost benets of thin lm electronics. [1][2][3][4][5][6][7][8] However, both the mass production of conventional OPV devices and many of their potential applications require exible substrates, which is a particular challenge for brittle electrode materials such as indium tin oxide (ITO). ITO is the most widely utilized transparent electrode material due to its relatively low sheet resistance on glass (15 U , À1 ) and high optical transparency (>90%) across relevant wavelengths, 9 however, due to the limited annealing conditions for polymeric substrates, the sheet resistance of ITO on plastic substrates is generally higher with values up to 40-60 U , À1 on poly-(ethylene terephthalate) (PET). Furthermore, ITO can fracture and delaminate from a exible plastic substrate when it is placed under strain. 10 Transparent electrodes suitable for R2R processing, which have the potential to overcome the shortcomings associated with ITO, typically utilize various metallic and semiconducting nanomaterials. Graphene, 5 single walled carbon nanotubes (SWCNTs), [11][12][13] ultra-thin metal lms, 14,15 metallic nanowires, [16][17][18][19] and composites of the aforementioned materials 14,[20][21][22][23] have all been used to create transparent electrodes for use in various organic optoelectronic devices. Typically, there is a trade-off between optical transparency and conductivity and hence a "gure of merit" is regularly used to compare transparent electrodes. This gure of merit utilises the sheet resistance of the electrode as well as the optical transmission at 550 nm, a larger value is an indication of better performance as a transparent conductor. 24 The gure of merit is derived from rearrangement of eqn (1) and shown in eqn (2), where T is the percent transmission at the wavelength of 500 nm and R sh is the sheet resistance in U , À1 .
The value of the gure of merit for ITO on glass is typically between 30 and 320 U À1 depending on processing condition and is approximately 290 U À1 for commercial supplies. 25 However, the gure of merit of ITO on exible substrates is around 40 U À1 resulting from a lower conductivity due the lower annealing temperatures required for the polymeric substrate.
Interwoven networks of silver nanowires (AgNW) and SWCNT have delivered sheet resistances as low as 4 U , À1 and optical transmission of greater than 85%, 26 however, this system was not planar and was not suitable for device fabrication. Reported planarization processes generally use rigid templates such as silicon or mica, which have naturally at surfaces that are 'transferred' to the electrode. [27][28][29] However, the rigid planar templates typically used are incompatible with R2R processing, which is an impediment for scaling production.
In this paper, we report planar interwoven AgNW:SWCNT:-PEDOT:PSS transparent electrodes fabricated via a scalable stamp R2R compatible technique. We describe the morphology of the electrode structure, the optoelectronic properties, and show that it performs competitively against ITO electrodes when used in OPV devices on solid and exible substrates.

Experimental
AgNWs were purchased from Seashell Technologies (San Diego, USA), which were supplied as a suspension (20.4 mg mL À1 ) in iso-propyl alcohol (IPA). The nanowires had a length of 5-50 mm with a diameter of approximately 100-200 nm. An aliquot of the AgNW suspension was diluted to 0.1 mg mL À1 with IPA (99.5% AR grade) and stored until use. Carboxylate functionalised (P3 type) SWCNTs with carbonaceous purity of >90% were purchased from Carbon Solutions (California, USA). The SWCNT bundle diameter was 5-15 nm and >50 mm in length. The carboxylate functionalised SWCNTs (50 mg) were puried by treatment with 3 M HNO 3 at reux for 12 h, followed by collection via vacuum ltration. It has been shown that mild acid treatment of SWCNTs improves dispersibility in water and also the performance of interwoven AgNW and SWCNT lms. 26,30 A suspension of the acid reuxed SWCNTs (12.5 mg) in deionised water (20 mL) was achieved using probe sonication (Sonics Vibracell™) at 40% amplitude for 2 min. The suspension was then diluted to a concentration of 0.25 mg mL À1 with deionised water.
AgNW:SWCNT interwoven networks were prepared via vacuum ltration through mixed cellulose ester membranes (MF-Millipore Membrane, USA, mixed cellulose esters, hydrophilic, 0.4 mm, 47 mm). Various volumes of the prepared AgNW (0.1 mg mL À1 ) and SWCNT (0.25 mg mL À1 ) solutions were added to deionised water (300 mL) so that a AgNW surface loading of 100 mg m À2 was produced in the nal nanocomposite electrode. Electrode patterning was achieved by placing a smaller pore size mixed cellulose ester template (MF-Millipore Membrane, mixed cellulose esters, hydrophilic, 0.025 mm, 47 mm) under the 0.4 mm membrane during ltration ( Fig. 1(a)). Aer ltration, the patterned electrodes were then placed on untreated poly(ethylene naphthalate) (PEN) ( Fig. 1(b)). The PEN and patterned electrodes were then passed through a laminator at 130 C ( Fig. 1(c)). The mixed cellulose ester lter paper was subsequently peeled away from the surface leaving behind the patterned AgNW:SWCNT nanocomposite on the surface of the PEN substrate. Subsequently, 100 mL of a 2 : 1 v/v PEDOT:PSS : IPA solution was spin-cast on top of the AgNW:SWCNT nanocomposite at 500 rpm for 5 s followed by 3000 rpm for 30 s. The PEDOT:PSS coated AgNW:SWCNT structure was then annealed at 140 C for 10 min ( Fig. 1(d)). 50 mL of Epotek 301 epoxy resin (transmission: T ¼ 99%) was then placed on top of the PEDOT:PSS coated AgNW:SWCNT structure. A PEN sheet with surface treatment for adhesion (Teonex lms, Teijin DuPont Films) was placed on top of the epoxy to create a PEN/AgNW:SWCNT:PEDOT:PSS/epoxy/PEN stack ( Fig. 1(e) and (f)). The stack was heated at 65 C for 1 h in an oven (Memmert, Germany) to cure the epoxy. The PEN used as the planar template (non-epoxy side) was peeled away to expose the planar active surface of the electrode.
Sheet resistance measurements were performed using the 4point probe technique (KeithLink Technology Co., Ltd., New Taipei City, Taiwan). The values reported are an average of 10 measurements on two separate 64 mm 2 samples. Flexibility testing was performed on a single planarized 25 mm 2 electrode. 2point probe measurements used during exibility investigations were performed by placing two conducting epoxy pads on the electrode at a separation distance of 20 mm. Transmission and haze was measured on samples (25 mm 2 ) using a Perkin-Elmer LAMBDA 950 UV/Vis/NIR Spectrophotometer with integrating sphere.
Scanning electron microscopy (SEM) images were acquired using a CamScan MX2500 (CamScan Optics, Cambridge, UK) working at an accelerating voltage of 10 kV and a working distance of 10 mm. The AgNW:SWCNT networks were stamp transferred onto a clean glass substrate and were imaged without a metal sputter coating to provide optimum contrast between the glass substrate, AgNWs and SWCNTs. Backscatter electron microscopy images were obtained at an accelerating voltage of 20 kV using an Inspect FEI F50 SEM with a eld emission gun source and concentric backscatter detector.
Topographical atomic force microscope (AFM) measurements were acquired using a Bruker Multimode AFM with a Nanoscope V controller. NSC15 Mikromasch Silicon tapping mode probes with a nominal spring constant of 40 N m À1 , resonant frequency of 325 kHz and tip radius equal to 10 nm were used. AFM images were acquired in tapping mode with all parameters including set-point, scan rate and feedback gains adjusted to optimize image quality and minimize imaging force. Roughness R q values were obtained from 10 Â 10 mm images aer 3 rd order plane tting. Conductivity of the planar AgNW:SWCNT:PEDOT:PSS electrodes were mapped using peak force tunnelling AFM (PF-TUNA) 31 on a Bruker Multimode AFM with a Nanoscope V controller. The soware used to acquire all AFM data was version 8.15. The cantilevers used to obtain the PF-TUNA images were Bruker PF-TUNA conducting probes with a spring constant of 0.4 N m À1 . The entire cantilever and tip is coated with 20 nm of each of platinum and iridium resulting in a total tip diameter of approximately 40 nm. PF-TUNA imaging parameters including set-point, scan rate, feedback gains, current sensitivity and applied bias were adjusted to optimize height and current image quality. The scanner was calibrated in the x, y, and z directions using silicon calibration grids (Bruker model numbers PG: 1 mm pitch, 110 nm depth and VGRP: 10 mm pitch, 180 nm depth).
Raman spectra and images were collected with a WiTEC alpha300R Microscope in confocal imaging Raman mode using a 100 Â (numerical aperture 0.9) objective with a 532 nm Nd-YAG green (E ¼ 2.33 eV) laser operating at constant power for each experiment. Laser power was kept at less than 10 mW during all measurements. Spectral images were acquired using an integration time of 5 s per pixel with each image composed of 50 pixels Â 50 pixels. Each pixel corresponds to a separate Raman spectrum, allowing thousands of spectra to be acquired during an image scan. Raman images were generated by selecting a region, in each spectra, in which a material specic peak is observed. The intensity of the selected region was plotted relative to the x, y position of the scanning laser. Single spectra were also acquired at points on the Raman images with typical integrations times between 30 s to 60 s and 2 to 3 accumulations per spectra. The spectral resolution is approximately 3 cm À1 . Raman data was collected by the WiTEC Control soware and analyzed using the WiTEC Project soware.

Device fabrication and testing
Organic solar cells with a structure of electrode/MoO x /poly(3-nhexylthiophene-2,5-diyl) (P3HT):phenyl-C61-butyric acid methyl ester (PCBM)/Al were fabricated with ITO or planar AgNW:SWCNT:PEDOT:PSS as the transparent electrode. Prepatterned ITO substrates (Xinyan Technology Ltd.) were cleaned using Alconox (Alconox) in de-ionized water. The ITO substrates were rinsed several times with de-ionized water and ultrasonicated in de-ionized water for 5 min. The ITO substrates were then ultra-sonicated in acetone (HPLC grade) and IPA (HPLC grade) for 10 min each, followed by drying under a ow of nitrogen. MoO x (Sigma Aldrich, Australia) (20 nm) was deposited using a vacuum thermal evaporator at a pressure  (4) 180 C, 2 min). Finally, a 100 nm layer of Al was deposited using a thermal evaporator at a pressure of $10 À6 mbar to complete the device structure. The devices fabricated had an active area of 0.2 cm 2 , which was dened using a shadow mask during cathode evaporation. Post fabrication, the devices were annealed on a hot plate at 180 C for 2 minutes.
Device characterization was performed using an Abet Triple-A (Abet Technologies) solar simulator. The solar mismatch of the xenon lamp (550 W Oriel) spectrum was minimized using an AM1.5G lter. Light intensity at $100 mW cm À2 AM1.5G was calibrated by using a National Renewable Energy Laboratory (NREL) certied standard silicon photodiode (2 cm 2 ), with a KG5 lter. A Keithley® 2400 source measurement unit in a two wire conguration set-up was used for current density-voltage (J-V) measurements. At least ve devices were characterized for each type of anode and for each fabrication condition. The short circuit current densities from J-V measurements of best devices were crossed checked with integrated current density from incident photon conversion efficiency (IPCE) measurements and they were within a difference of 10%. IPCE spectra were taken using a PV Measurement QEX7 setup, which was calibrated with an NREL certied photodiode and operated without white light bias and chopped and locked in the small perturbation limit.
All fabrication steps and testing of devices were done in a class 1000 cleanroom. Except for cleaning of the ITO substrates, all other steps for device fabrication and testing were done in an inert environment (MBraun glove box, O 2 < 0.1 ppm; H 2 O < 0.1 ppm). The evaporations were carried out using a thermal evaporator connected to a glove box.

Results and discussion
Planar electrode fabrication process Planar AgNW:SWCNT:PEDOT:PSS electrodes were fabricated via a scalable, R2R compatible, solution process using a li off technique from a exible substrate. Processes reported previously have relied upon the dissolution of a mixed cellulose ester membrane to transfer the AgNW:SWCNT network to a rigid substrate. 29 In the process presented in Fig. 1, the relative adhesion of the AgNW:SWCNT network to various substrates was investigated and it was found that the interwoven network could be efficiently stamp transferred to rigid or exible substrates when SWCNTs were co-deposited with AgNWs ( Fig. 1(d)). Aer preparing a mixed dispersion of AgNWs and SWCNTs in de-ionized water, the process consisted of the following basic steps: (a) ltration through a masked cellulose ester membrane to create the desired interwoven pattern; (b) placing the cellulose ester lter on top of the planar PEN template; (c) passing the cellulose ester membrane/ AgNW:SWCNT/PEN stack through heated rollers; (d) peeling away the cellulose ester membrane; (e) coating the PEN substrate and the interwoven electrode with PEDOT:PSS and drying; (f) addition of an epoxy base layer to adhere and transfer to another glass/plastic substrate and (g) peeling away the planar PEN template to reveal the electrode active surface. For a more detailed description of please see methods section.
SWCNTs have previously been shown to wrap around the AgNWs during solution phase processing to create an interwoven self-supporting network. 26 It is the strength of the interaction between AgNWs and SWCNTs that enables AgNW stamp transfer (Fig. 1(c)) to a substrate to occur, and which give its important properties. Fig. 2 shows a comparison of the sheet resistance of an electrode formed by the AgNW stamp transfer method with and without SWCNTs on glass. Measurements were taken aer removal of the cellulose ester membrane (step (d), Fig. 1).
The AgNW only lm (Fig. 2, diamonds) produced a conducting network with a sheet resistance of approximately 30-50 U , À1 with a standard deviation of up to 30.8 U , À1 . This is signicantly poorer than the sheet resistance of commercial ITO on glass (10-15 U , À1 ) although it is similar to ITO on exible substrates. 32 Nevertheless, the sheet resistance is too high for large area devices. The error associated with each sample is large and reects the poor reproducibility of the technique for AgNW only electrodes. The poor transfer is presumably due to weak attractive forces between the AgNW network and the glass substrate during lamination.
SEM was conducted on the stamp transferred AgNW:SWCNT networks from mixed cellulose ester lter papers (0.4 mm) containing 109 mg m À2 loading with and without SWCNTs, shown in Fig. 3(a) and (b), respectively. In Fig. 3(a) it can be seen that without the presence of SWCNTs the AgNW forms a dense network aer stamp transfer from the cellulose ester membrane. However, the sheet resistance (34.7 AE 10 U , À1 ) suggests that the network is not well interconnected aer stamp transfer (Fig. 1, step (d)). Fig. 3(b) shows a very high density of co-deposited AgNWs and SWCNTs on a the glass surface. The high resolution inset in Fig. 3(b) shows that every SWCNT is closely associated with the AgNW network leading to a sheet resistance signicantly lower than that of the AgNW only network at 15.2 AE 1 U , À1 . Fig. 3(c) and (d) are high magnication PF-TUNA images showing the height (Fig. 3(c)) and conductivity (Fig. 3(d)) of the AgNW:SWCNT electrode. Importantly, we can see from Fig. 3(d) Fig. 2 Sheet resistance of AgNW networks transferred from cellulose ester filter paper to PEN (Fig. 1, step (d)) without SWCNTs (diamonds) and with 20 wt% SWCNTs (squares). Values are an average of 20 measurements on 3 separate 8 Â 8 mm electrodes. Error bars represent 1 standard deviation. that the SWCNTs and the AgNWs are electrically connected, which is benecial for charge collection at the active surface of the electrode upon planarization.

Planar electrode properties
Upon completion of the planarization process (Fig. 1, step (g)) the active surface of the electrode had a sheet resistance of 4-7 U , À1 and an average transmission of >86% over wavelength range of 800-400 nm and a haze value of 11.6% over the same wavelength range (Fig. 4(c)). Importantly, the gure of merit is the same for electrodes produced on glass substrates as well as plastic lm using the process outlined in Fig. 1 with values between approximately 344-400 U À1 which compares with between 48 and 208 U À1 for AgNW composites reported in the literature. The improvement over previous work (Table 1) is due to the reduction of fabrication steps by stamp processing the AgNW:SWCNT network and the ability to process on a PEN planar template instead of silicon.
Evidence for the AgNW:SWCNT:PEDOT:PSS structure being present at the active surface of the electrode is shown in Fig. 4(a) and a cross-section of the electrode structure is shown in Fig. 4(b). The roughness of the electrode is 4.78 nm over a 10 Â 10 mm plane tted image. The cross-section image (Fig. 4(b)) shows that the overall thickness of the epoxy layer is approximately 5 mm.
Importantly, Fig. 4(b) shows that the AgNWs are present at the active layer and not found distributed throughout the epoxy.
Raman microscopy of the electrode surface revealed the position of PEDOT:PSS and SWCNTs relative to the AgNWs present at the surface of the electrode. An example Raman spectrum is shown in Fig. 5(d) and other spectra are included in the ESI. † In the optical image of Fig. 5(a) the silver nanowires are easily identied. However, the other components of the electrode namely, PEDOT:PSS and SWCNTs are not visible optically. By plotting the intensity of the Raman signal at 440 cm À1 , which corresponds to the sulphonate bending from PSS, 39 and 1597 cm À1 , corresponding to the G band for SWCNTs, 40 the regions on the electrode surface that contained PEDOT:PSS and SWCNTs, respectively can be identied. Fig. 5(b) shows the regions of PEDOT:PSS on the active surface of the electrode and indicates that PEDOT:PSS is present over the entire surface. Fig. 5(b) shows that the PEDOT:PSS signal is enhanced when in close association with AgNWs. This observation has two possible explanations. Firstly, since the PEDOT:PSS is spin-cast onto the AgNW network it may become trapped at the apex of adjoining nanowires and the planar PEN template resulting in a high concentration of PEDOT:PSS relative to the areas between the wires. Alternatively, it may be due to enhancement of the Raman signal at the surface of AgNWs. 41,42 It is apparent that the SWCNTs on the surface of the electrode are present on the AgNWs as well as in between   the nanowire network. It is unsurprising that there are significant quantities of SWCNTs in close association with the AgNWs since the solution phase deposition has been seen to encourage this interaction. 26 The SWCNTs, which are spread out between the AgNW network are expected to increase the charge collecting capability of the electrode and also serve to connect and AgNW that are separate from the primary conducting network. Fig. 5(d) is single point Raman spectra collected at positions 1 and 2 as indicated on the optical image Fig. 5(a). The Raman spectra at position 1 shows a strong Raman signal at 1597 cm À1 , which is assigned to the G-band typical for SWCNTs. A peak at 2680 cm À1 can also be observed which is characteristic of the 2D region for SWCNTs. 40 The Raman spectra at position 2 does not show any of the characteristic Raman peaks for SWCNTs but instead shows Raman peaks typical for PEDOT:PSS. 39,43,44 The planar AgNW:SWCNT:PEDOT:PSS electrodes on PEN were found to retain a low sheet resistance aer bending at a radius of 5 mm up to 500 times (Fig. 6). Aer each bend cycle the sheet resistance of the electrode was measured using a fourpoint probe and from two conductive epoxy pads on the electrode at a separation distance of 20 mm. The open circles in Fig. 6 shows that the four-point probe measurement increases slightly over the 500 bend cycles, however, small indentations were le in the so polymer electrode lm aer each measurement and are believed to contribute to this rise. To overcome this problem two conductive epoxy pads were placed on the electrode and used as stable contact points for resistance measurements using a multimeter (Fig. 6, open squares). The measurement shows that over the 500 bend cycles at a 5 mm bend radius there is no signicant change in the resistivity of the electrode. The high exibility of the planar AgNW:SWCNT:PEDOT:PSS/PEN electrode makes is suitable for R2R manufacturing processes and a range of organic optoelectronic applications.

OPV devices
The current density (J) and voltage (V) characteristics of the best devices fabricated on AgNW:SWCNT:PEDOT:PSS/glass and comparative ITO electrodes are shown in Fig. 7(a) and (b), respectively with the device statistics for each structure and annealing conditions given in Table 2. For the AgNW:SWCNT:PEDOT:PSS on glass electrodes, annealing for 10 min at 120 C resulted in the best device performance, with a PCE of 1.85 AE 0.18%. For the same annealing condition the average ITO device efficiency was 2.2 AE 0.1%. On average, the devices fabricated with AgNW:SWCNT:PEDOT:PSS electrodes achieved $86% of the efficiency of comparable OPV devices fabricated with an ITO transparent electrode. Further improvements are expected once device fabrication parameters are fully explored.
An annealing temperature of 120 C, which gave the best device performance on glass, is compatible with the use of exible plastic substrates. A device fabricated on a exible PEN substrate produced an efficiency of 1.2%, as shown in Fig. 8. This is the rst time that a R2R compatible nanocomposite electrode has been used for fabrication of a exible ITO device.

Conclusions
A roll to roll (R2R) compatible, solution process for the production of highly conductive, transparent and planar electrode based on silver nanowires (AgNW) and single walled carbon nanotubes (SWCNT) imbedded into PEDOT:PSS has been demonstrated, with a sheet resistance of 4-7 U , À1 and a transmission of >86% between 800 and 400 nm. The gure of merit of 344-400 U À1 is the highest published for AgNW composite electrodes.
The low temperature, solution based process has been demonstrated to be able to produce electrodes with the same properties of both glass and exible plastic (PEN) substrates with OPVs fabricated on the electrodes delivering $90% of the performance of equivalent ITO based devices on a glass substrate, this is without fabrication optimisation for the exible electrode.
The process to fabricate the high gure of merit electrodes uses the entanglement of silver nanowires and single walled carbon nanotubes to provide sufficient mechanical stability to enable the transfer from a cellulose ester membrane.
The planar electrode surface structure exists as an interpenetrating network of AgNWs and SWCNTs imbedded in PEDOT:PSS with substantial concentrations of SWCNTs and PEDOT:PSS observed at the electrically active surface. In addition to active participation in charge collection and transport, the SWCNTs have also been demonstrated to facilitate the stamp transfer from a cellulose ester membrane to a receiving substrate by providing mechanical stability to the mesh with signicantly improved consistency of transfer compared to a AgNW only mesh.
The planar AgNW:SWCNT:PEDOT:PSS electrode is highly exible and resistance to fatigue with no degradation of the conductivity aer bending at a radius of 5 mm up to 500 times opening the potential for use in novel applications and R2R fabrication of organic optoelectronic devices.